Re: Intentional Corruption
Date: Thu, 14 Feb 2008 12:21:39 -0500
Disclaimer: I'm not a lawyer and this is not a legal advice 8-) And please do not do this on your production system ;-)
Described approach should work if your datafiles are located on the filesystem and your system is a UNIX one
- calculate the offset in bytes for your table in the datafile by using dba_extents view and the fact that size of any datafile is one database block over the size reported in v$datafile (alternatively you can use DBMS_ROWID to corrupt a specific row)
- dd if=/dev/zero of=<datafile> bs=1 seek=<calculated offset> count=<number of bytes to corrupt> conv=notrunc (this syntax is for Solaris)
- in shell sync; sync; sync to make sure filesystem cache was flushed to disk.
- flush your db cache by using alter system flush buffer_cache
- do a full table scan
That should be enough to detect a corruption.
PS: you can experiment with <number of bytes to corrupt> or <calculated offset> to get desired result. Keep in mind that you should aim to corrupt the blocks below HWM of the target table in order for FTS to detect a corruption.
On 2/14/08, Ahbaid Gaffoor <ahbaid_at_att.net> wrote:
> Is there a way to intentionally introduce block level corruption?
> I'd like to play with some scenarios to learn more about DBMS_REPAIR